CTOs buying Mac Studios for local AI hit the same wall: models crash, memory runs out, switching is slow, nothing is observable, and they can't trust the stack for real work. But even if the models run perfectly, they still have to figure out how to actually use them — routing workloads, building agent logic, connecting to their systems, keeping sessions alive, handling failures. A reliable local endpoint isn't enough. They need the whole stack working together.
Foundry is a complete private AI infrastructure stack for Apple Silicon. Not just local inference — inference, orchestration, messaging, and observability, all working together out of the box.
Three offers:
Foundry Advisory (£299 one-time)
We assess your hardware, workloads, and current AI spend. We tell you which Mac to buy, which models to run, which workloads to move local, and which to keep on cloud. You get a written recommendation pack and setup scripts.
Foundry Managed (£999 setup + £99/mo)
We set up your Mac Studio as a complete private AI infrastructure node:
Foundry On-Prem (£2–5k/mo)
For compliance-sensitive teams. Everything in Managed, plus audit logging, no-cloud mode, a support contract, and a runbook your risk team can read. Your data never leaves your network.
| Layer | What it does | What the CTO sees |
|---|---|---|
| **Foundry** | Model management, capacity planning, health checks, benchmarks | "Which models are running? Are they healthy? What fits in memory?" |
| **OpenClaw** (or Hermes) | Agent orchestration **OR** messaging integration | "My support queries get handled automatically" / "The AI reads our Slack and responds" |
| **llm_stats** | Real-time observability — endpoints, memory, activity, crash risk | "I can see at a glance that everything is working." |
The harness layer is one or the other, not both. A CTO chooses:
Combining both is possible for advanced setups but is not the default configuration.
You don't need to understand agentic architecture. You tell us:
We configure the full stack — models, agents, integrations, monitoring — and hand you a working system. Your team interacts with it through your existing tools (Slack, email, API). The infrastructure is invisible.
We're looking for 3 design partners — CTOs who want to cut their AI bill and run private AI on their own hardware. You tell us what workloads you want local. We make the whole stack work.